Midjourney’s Whole Body Ultrasound System: A Radiologist’s Thoughts
Earlier this month, the AI technology company Midjourney announced a prototype whole body ultrasound scanner that can reportedly scan a patient in 60 seconds. Because the system uses ultrasound, there is no ionizing radiation. Their announcement prompted a flurry of commentary from tech aficionados and physicians alike, with some enthusiasts declaring this to be a “game changer” for medical care and others dismissing this as hype.
As a diagnostic radiologist with 30-plus years of experience interpreting ultrasound, CT, and MRI scans, I’d like to offer a few thoughts.
1) The underlying technology is pretty innovative.
The patient is immersed in a water bath, and the system scans the patient with a 360-degree ring of ultrasound scanners fully surrounding the patient. The software analyzes the data from the sensors and generates a series of image “slices” much like a CT scan. The 360-degree view allows the system to generate a more detailed image than is possible with a standard ultrasound scanner which views the patient from just a single angle. This is similar to a CT scanner which shows a 360-view of the patient that is superior than the single image of a standard x-ray. However, because the Midjourney system utilizes ultrasound, there is no radiation exposure (as opposed to with CT scans.) The Midjourney system is also cheaper than a whole-body MRI.
2) The technology could be pretty useful for the stated goal of body composition analysis.
After the Midjourney system acquires the images, their AI tool does a pretty good job of determining the boundaries of various internal organs from body fat (“segmentation”). This allows it to calculate informaton such as the percentage of muscle mass or body fat within the torso or extremities. This data can be helpful in tracking changes in body composition due to alterations in patient diet, exercise level, or drug treatments.
3) The ultrasound technology cannot see some critical body structures.
One of the basic limits of conventional ultrasound is that it does not penetrate bone. Nor does it penetrate the interface between tissue and air. (This is why the patient must be submerged in a water bath.) This means that the ultrasound images cannot generate images of the brain (which is surrounded by the skull) or the lungs. It does generate credible images of solid organs in the abdomen, such as the liver, spleen, kidneys, pancreas, etc.
4) The system is not FDA approved for screening of early cancers.
The Midjourney company does not promote their device for routine screening of early cancers, although some online enthusiasts earnestly hope this will be the case soon. Based on the images provided by Midjourney, I would not feel confident in determining whether or not there was a early tumor (less than 1 cm in size) in organs such as the liver or kidneys. The image quality and resolution of the Midjourney images I’ve seen are not yet up to that of traditional ultrasound.
5) If the image quality improves, one potentially useful application would be close serial monitoring of patient anatomical structures.
For example, a small lesion in the kidney could be monitored on, say, a weekly or monthly basis to see if it grows. Or a known tumor can be monitored closely to see how it responds to chemotherapy. Asymptomatic lesions discovered incidentally in populations could also be monitored en masse to generate a robust database that would help researchers learn which types of lesions are more (or less) likely to turn malignant. The combination of relatively low cost and absence of radiation makes this a much more feasible research endeavor than monitoring large numbers of patients with serial CT or MRI scans in organs amenable to ultrasound visualization.
6) Too much of the acrimonious online conversation about Midjourney between “tech bros” and physicians is based on unfortunate stereotypes.
Not all physicians are Luddites who refuse to embrace new technologies. Some physicians skeptical of using Midjourney for widespread cancer screening have raised legitimate questions about whether radiologists can interpret potentially vast new data sets from scanners that are not yet FDA approved. Plus there are genuine concerns about whether any system of whole-body imaging for early detection of cancers in asymptomatic patients will lead to harm from unnecessary biopsy procedures with known risks and complication rates.
On the other hand, not all technology enthusiasts are shameless “grifters” trying to make a quick buck from gullible affluent healthy patients fearful about early cancers (the so-called “worried well.”) Many tech innovators genuinely wish to develop tools and technologies that will enhance human life.
I welcome any new medical technologies that can improve patient care. I especially love innovations in medical imaging that could allow physicians to make important diagnoses faster, cheaper, more accurately, and/or more safely. But any marketing claims made about potential benefits of a new health technology need to be validated through an appropriate process of objective technical and medical evaluation. I hope technology enthusiasts and physician can collaborate towards this goal in a fashion that generates genuine value for everyone.
